Revert of De-proc Color32 (patchset #5 id:80001 of https://codereview.chromium.org/1104183004/)

Reason for revert:
duh

Original issue's description:
> De-proc Color32
>
> Also strips SK_SUPPORT_LEGACY_COLOR32_MATH,
> which is no longer needed.
>
> Seems handy to have SkTypes include the relevant intrinsics when
> we know we've got them, but I'm not married to it.
>
> Locally this looks like a pointlessly small perf win, but I'm mostly
> keen to get all the code together.
>
> BUG=skia:
>
> Committed: https://skia.googlesource.com/skia/+/376e9bc206b69d9190f38dfebb132a8769bbd72b
>
> Committed: https://skia.googlesource.com/skia/+/d65dc0cedd5b50dd407b6ff8fdc39123f11511cc

TBR=reed@google.com,mtklein@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Review URL: https://codereview.chromium.org/1102363006

1 files changed, 65 insertions, 0 deletions

diff --git a/src/opts/SkBlitRow_opts_SSE2.cpp b/src/opts/SkBlitRow_opts_SSE2.cpp
index 7f5b6779cf..59375f1831 100644
--- a/src/opts/SkBlitRow_opts_SSE2.cpp
+++ b/src/opts/SkBlitRow_opts_SSE2.cpp
@@ -232,6 +232,71 @@ void S32A_Blend_BlitRow32_SSE2(SkPMColor* SK_RESTRICT dst,
     }
 }
 
+#define SK_SUPPORT_LEGACY_COLOR32_MATHx
+
+/* SSE2 version of Color32()
+ * portable version is in core/SkBlitRow_D32.cpp
+ */
+// Color32 and its SIMD specializations use the blend_256_round_alt algorithm
+// from tests/BlendTest.cpp.  It's not quite perfect, but it's never wrong in the
+// interesting edge cases, and it's quite a bit faster than blend_perfect.
+//
+// blend_256_round_alt is our currently blessed algorithm.  Please use it or an analogous one.
+void Color32_SSE2(SkPMColor dst[], const SkPMColor src[], int count, SkPMColor color) {
+    switch (SkGetPackedA32(color)) {
+        case   0: memmove(dst, src, count * sizeof(SkPMColor)); return;
+        case 255: sk_memset32(dst, color, count);               return;
+    }
+
+    __m128i colorHigh = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_set1_epi32(color));
+#ifdef SK_SUPPORT_LEGACY_COLOR32_MATH  // blend_256_plus1_trunc, busted
+    __m128i colorAndRound = colorHigh;
+#else                          // blend_256_round_alt, good
+    __m128i colorAndRound = _mm_add_epi16(colorHigh, _mm_set1_epi16(128));
+#endif
+
+    unsigned invA = 255 - SkGetPackedA32(color);
+#ifdef SK_SUPPORT_LEGACY_COLOR32_MATH  // blend_256_plus1_trunc, busted
+    __m128i invA16 = _mm_set1_epi16(invA);
+#else                          // blend_256_round_alt, good
+    SkASSERT(invA + (invA >> 7) < 256);  // We should still fit in the low byte here.
+    __m128i invA16 = _mm_set1_epi16(invA + (invA >> 7));
+#endif
+
+    // Does the core work of blending color onto 4 pixels, returning the resulting 4 pixels.
+    auto kernel = [&](const __m128i& src4) -> __m128i {
+        __m128i lo = _mm_mullo_epi16(invA16, _mm_unpacklo_epi8(src4, _mm_setzero_si128())),
+                hi = _mm_mullo_epi16(invA16, _mm_unpackhi_epi8(src4, _mm_setzero_si128()));
+        return _mm_packus_epi16(_mm_srli_epi16(_mm_add_epi16(colorAndRound, lo), 8),
+                                _mm_srli_epi16(_mm_add_epi16(colorAndRound, hi), 8));
+    };
+
+    while (count >= 8) {
+        __m128i dst0 = kernel(_mm_loadu_si128((const __m128i*)(src+0))),
+                dst4 = kernel(_mm_loadu_si128((const __m128i*)(src+4)));
+        _mm_storeu_si128((__m128i*)(dst+0), dst0);
+        _mm_storeu_si128((__m128i*)(dst+4), dst4);
+        src   += 8;
+        dst   += 8;
+        count -= 8;
+    }
+    if (count >= 4) {
+        _mm_storeu_si128((__m128i*)dst, kernel(_mm_loadu_si128((const __m128i*)src)));
+        src   += 4;
+        dst   += 4;
+        count -= 4;
+    }
+    if (count >= 2) {
+        _mm_storel_epi64((__m128i*)dst, kernel(_mm_loadl_epi64((const __m128i*)src)));
+        src   += 2;
+        dst   += 2;
+        count -= 2;
+    }
+    if (count >= 1) {
+        *dst = _mm_cvtsi128_si32(kernel(_mm_cvtsi32_si128(*src)));
+    }
+}
+
 void Color32A_D565_SSE2(uint16_t dst[], SkPMColor src, int count, int x, int y) {
     SkASSERT(count > 0);